Abstract: Standard machine translation typically requires waiting for the complete source language input before beginning the translation. This offline interaction model falls short in meetin g the demands of real-time scenarios, such as simultaneous interpretation, international c onferences, and national travel. In contrast, simultaneous translation necessitates that the model generates target outputs while receiving the source inputs, thereby achieving tran slation under low latency. As a result, it has garnered increasing attention from researchers. However, simultaneous translation faces several challenges, primarily due to the need t o determine the appropriate moment to begin translating and to generate translations b ased on incomplete inputs. Recently, much research has focused on developing low-laten cy, high-quality simultaneous translation models, yet there remains a lack of a comprehe nsive review of existing methods and research trends. This paper not only provides a deta iled classification and analysis of current simultaneous translation methods from the pers pective of application scenarios but also forecasts future research directions. Furthermore, this paper systematically organizes relevant literature, offering researchers a valuable res ource to quickly grasp the latest advancements and emerging trends in the field of simult aneous translation.